Inferendum

Inferendum is a deterministic causal analysis engine built on EconML and DoWhy that estimates causal effects from experimental or observational data. It presents naive correlational results alongside causal estimates, identifies heterogeneous effects with multiple-testing corrections, and provides refutation tests and sensitivity analyses. The tool supports randomised A/B tests and observational studies adjusted with doubly-robust machine learning, offering a reproducibility pack with LaTeX tables for every run.

The estimation engine is plain, deterministic Python with no large language model touching the numbers, ensuring same data in yields same answer out. Before applying the method to user data, Inferendum offers validation against a synthetic twin with a known ground-truth effect to ensure reliability.

Inferendum provides a plain-language assumptions checklist on every result, alongside refutation tests and sensitivity analyses to help users understand the conditions under which the causal estimates hold. It also shows the gap between naive correlational headlines and the actual causal estimate side by side.

The tool covers two main study designs: randomised experiments like RCTs or A-B tests where treatment is randomised, and observational studies where it adjusts for measured covariates using doubly-robust machine learning. Additional designs like difference-in-differences and regression discontinuity are on the roadmap.

Inferendum is built and maintained by Epon Khalil, who holds an MSc in Financial Economics from the University of Glasgow. His research focuses on causal inference and the evaluation of digital inclusion policy, which forms the basis of the tool's methods.

Users can get in touch with the team via the business e-mail address [email protected] for questions, collaborations, or to try a dataset. The tool is designed for research and real decisions, catering to those needing transparent causal inference.

The platform is located in the UK, as indicated by the founder's affiliation with the University of Glasgow and the use of a UK-based e-mail domain. It serves researchers and decision-makers globally who need deterministic causal analysis.

The tool offers a synthetic-twin calibration feature that validates the method against a simulated twin with a known true effect before trusting the results on actual data. This step ensures the method works properly before application.

Inferendum includes a targeting rule that identifies who the effect is for, providing honest, multiple-testing-corrected subgroups to help with decision-making. This feature addresses heterogeneous effects rather than assuming a uniform impact.

The website states that the estimation engine contains no large language model, distinguishing it from other AI-driven tools that may dress up correlation as cause. This deterministic approach ensures consistency in results.

The tool ships with a reproducibility pack that includes copy-paste LaTeX tables, allowing users to easily document and share their findings. Every run provides this pack to ensure transparency.

The roadmap includes plans to add more designs such as difference-in-differences, synthetic control, regression discontinuity, and instrumental variables, each validated against synthetic ground truth before release.

The official website of Inferendum is inferendum.com.

Inferendum has no reviews on Trustburn yet. Anyone who has used it can write the first review on its Trustburn page.